iterator: combine fs+workdir iterators more completely

Drop some of the layers of indirection between the workdir and the
filesystem iterators.  This makes the code a little bit easier to
follow, and reduces the number of unnecessary allocations a bit as
well.  (Prior to this, when we filter entries, we would allocate them,
filter them and then free them; now we do the filtering before
allocation.)

Also, rename `git_iterator_advance_over_with_status` to just
`git_iterator_advance_over`.  Mostly because it's a fucking long-ass
function name otherwise.

src/checkout.c

@@ -394,7 +394,7 @@ static int checkout_action_wd_only(
 		git_buf_sets(&data->tmp, wd->path);
 		saved_wd.path = data->tmp.ptr;
 
-		error = git_iterator_advance_over_with_status(
+		error = git_iterator_advance_over(
 			wditem, &untracked_state, workdir);
 		if (error == GIT_ITEROVER)
 			over = true;
@@ -930,7 +930,7 @@ static int checkout_conflicts_load(checkout_data *data, git_iterator *workdir, g
 	git_index *index;
 
 	/* Only write conficts from sources that have them: indexes. */
-	if ((index = git_iterator_get_index(data->target)) == NULL)
+	if ((index = git_iterator_index(data->target)) == NULL)
 		return 0;
 
 	data->update_conflicts._cmp = checkout_conflictdata_cmp;
@@ -1080,7 +1080,7 @@ static int checkout_conflicts_coalesce_renames(
 	size_t i, names;
 	int error = 0;
 
-	if ((index = git_iterator_get_index(data->target)) == NULL)
+	if ((index = git_iterator_index(data->target)) == NULL)
 		return 0;
 
 	/* Juggle entries based on renames */
@@ -1138,7 +1138,7 @@ static int checkout_conflicts_mark_directoryfile(
 	const char *path;
 	int prefixed, error = 0;
 
-	if ((index = git_iterator_get_index(data->target)) == NULL)
+	if ((index = git_iterator_index(data->target)) == NULL)
 		return 0;
 
 	len = git_index_entrycount(index);
@@ -2378,7 +2378,7 @@ static int checkout_data_init(
 		if ((error = git_repository_index(&data->index, data->repo)) < 0)
 			goto cleanup;
 
-		if (data->index != git_iterator_get_index(target)) {
+		if (data->index != git_iterator_index(target)) {
 			if ((error = git_index_read(data->index, true)) < 0)
 				goto cleanup;
 
@@ -2600,7 +2600,7 @@ int git_checkout_iterator(
 		(error = checkout_create_conflicts(&data)) < 0)
 		goto cleanup;
 
-	if (data.index != git_iterator_get_index(target) &&
+	if (data.index != git_iterator_index(target) &&
 		(error = checkout_extensions_update_index(&data)) < 0)
 		goto cleanup;
 

src/diff.c

@@ -826,8 +826,7 @@ static int maybe_modified(
 	 */
 	} else if (git_oid_iszero(&nitem->id) && new_is_workdir) {
 		bool use_ctime = ((diff->diffcaps & GIT_DIFFCAPS_TRUST_CTIME) != 0);
-		git_index *index;
-		git_iterator_index(&index, info->new_iter);
+		git_index *index = git_iterator_index(info->new_iter);
 
 		status = GIT_DELTA_UNMODIFIED;
 
@@ -980,15 +979,14 @@ static int iterator_advance_into(
 	return error;
 }
 
-static int iterator_advance_over_with_status(
+static int iterator_advance_over(
 	const git_index_entry **entry,
 	git_iterator_status_t *status,
 	git_iterator *iterator)
 {
-	int error;
+	int error = git_iterator_advance_over(entry, status, iterator);
 
-	if ((error = git_iterator_advance_over_with_status(
-			entry, status, iterator)) == GIT_ITEROVER) {
+	if (error == GIT_ITEROVER) {
 		*entry = NULL;
 		error = 0;
 	}
@@ -1056,7 +1054,7 @@ static int handle_unmatched_new_item(
 				return iterator_advance(&info->nitem, info->new_iter);
 
 			/* iterate into dir looking for an actual untracked file */
-			if ((error = iterator_advance_over_with_status(
+			if ((error = iterator_advance_over(
 					&info->nitem, &untracked_state, info->new_iter)) < 0)
 				return error;
 

src/iterator.h

@@ -34,10 +34,19 @@ typedef enum {
 	GIT_ITERATOR_DONT_AUTOEXPAND  = (1u << 3),
 	/** convert precomposed unicode to decomposed unicode */
 	GIT_ITERATOR_PRECOMPOSE_UNICODE = (1u << 4),
+	/** never convert precomposed unicode to decomposed unicode */
+	GIT_ITERATOR_DONT_PRECOMPOSE_UNICODE = (1u << 5),
 	/** include conflicts */
-	GIT_ITERATOR_INCLUDE_CONFLICTS = (1u << 5),
+	GIT_ITERATOR_INCLUDE_CONFLICTS = (1u << 6),
 } git_iterator_flag_t;
 
+typedef enum {
+	GIT_ITERATOR_STATUS_NORMAL = 0,
+	GIT_ITERATOR_STATUS_IGNORED = 1,
+	GIT_ITERATOR_STATUS_EMPTY = 2,
+	GIT_ITERATOR_STATUS_FILTERED = 3
+} git_iterator_status_t;
+
 typedef struct {
 	const char *start;
 	const char *end;
@@ -57,6 +66,8 @@ typedef struct {
 	int (*current)(const git_index_entry **, git_iterator *);
 	int (*advance)(const git_index_entry **, git_iterator *);
 	int (*advance_into)(const git_index_entry **, git_iterator *);
+	int (*advance_over)(
+		const git_index_entry **, git_iterator_status_t *, git_iterator *);
 	int (*reset)(git_iterator *);
 	int (*reset_range)(git_iterator *, const char *start, const char *end);
 	int (*at_end)(git_iterator *);
@@ -67,8 +78,13 @@ struct git_iterator {
 	git_iterator_type_t type;
 	git_iterator_callbacks *cb;
 	git_repository *repo;
+
 	char *start;
+	size_t start_len;
+
 	char *end;
+	size_t end_len;
+
 	bool started;
 	bool ended;
 	git_vector pathlist;
@@ -76,6 +92,7 @@ struct git_iterator {
 	int (*strcomp)(const char *a, const char *b);
 	int (*strncomp)(const char *a, const char *b, size_t n);
 	int (*prefixcomp)(const char *str, const char *prefix);
+	int (*entry_srch)(const void *key, const void *array_member);
 	size_t stat_calls;
 	unsigned int flags;
 };
@@ -183,6 +200,28 @@ GIT_INLINE(int) git_iterator_advance_into(
 	return iter->cb->advance_into(entry, iter);
 }
 
+/* Advance over a directory and check if it contains no files or just
+ * ignored files.
+ *
+ * In a tree or the index, all directories will contain files, but in the
+ * working directory it is possible to have an empty directory tree or a
+ * tree that only contains ignored files.  Many Git operations treat these
+ * cases specially.  This advances over a directory (presumably an
+ * untracked directory) but checks during the scan if there are any files
+ * and any non-ignored files.
+ */
+GIT_INLINE(int) git_iterator_advance_over(
+	const git_index_entry **entry,
+	git_iterator_status_t *status,
+	git_iterator *iter)
+{
+	if (iter->cb->advance_over)
+		return iter->cb->advance_over(entry, status, iter);
+
+	*status = GIT_ITERATOR_STATUS_NORMAL;
+	return git_iterator_advance(entry, iter);
+}
+
 /**
  * Advance into a tree or skip over it if it is empty.
  *
@@ -273,35 +312,12 @@ extern int git_iterator_cmp(
 extern int git_iterator_current_workdir_path(
 	git_buf **path, git_iterator *iter);
 
-/* Return index pointer if index iterator, else NULL */
-extern git_index *git_iterator_get_index(git_iterator *iter);
-
-typedef enum {
-	GIT_ITERATOR_STATUS_NORMAL = 0,
-	GIT_ITERATOR_STATUS_IGNORED = 1,
-	GIT_ITERATOR_STATUS_EMPTY = 2,
-	GIT_ITERATOR_STATUS_FILTERED = 3
-} git_iterator_status_t;
-
-/* Advance over a directory and check if it contains no files or just
- * ignored files.
- *
- * In a tree or the index, all directories will contain files, but in the
- * working directory it is possible to have an empty directory tree or a
- * tree that only contains ignored files.  Many Git operations treat these
- * cases specially.  This advances over a directory (presumably an
- * untracked directory) but checks during the scan if there are any files
- * and any non-ignored files.
- */
-extern int git_iterator_advance_over_with_status(
-	const git_index_entry **entry, git_iterator_status_t *status, git_iterator *iter);
-
 /**
  * Retrieve the index stored in the iterator.
  *
- * Only implemented for the workdir iterator
+ * Only implemented for the workdir and index iterators.
  */
-extern int git_iterator_index(git_index **out, git_iterator *iter);
+extern git_index *git_iterator_index(git_iterator *iter);
 
 typedef int (*git_iterator_walk_cb)(
 	const git_index_entry **entries,